Tasty Watermelon Identifier

Inspiration

We took notice that many individuals don't know how to determine the quality of watermelon through just looks, so we decide to create an app to assist with this.

What it does

By taking pictures on an Android phone, the app identifies whether or not the object is indeed a watermelon, before using a custom made model to determine the quality of watermelon, that is later presented on screen for the viewer to see

How we built it

We used a variety of different IDES, as well as software to assist with the app design which include:

• Pycharm
• Pytorch
• Android Studio
• ImageNet
• Kaggle
• Roboflow

Challenges we ran into

The biggest issue we faced was making the custom model compatible with Android Studio. The incompatibility between being unable to convert Googlenet tuples to Tensor was especially challenging. We also faced troubles with being able to properly work on the same project at the same time, as Github was acting weird during the Hackathon.

Accomplishments that we're proud of

We're most proud of being able to successfully train our custom model, make it communicate with bitmaps of images taken by the phone on the Android end, and being able to successfully identify watermelons from other objects, as well as quality of said watermelon.

What we learned

We learnt how to utilize pytorch and machine learning to help us identify watermelons

What's next for Tasty Watermelon Identifier

We hope that we can further improve this app's accuracy and the ability for the user to select different fruits for image identification.

Built With

• android-studio
• imagenet
• java
• python
• pytorch
• roboflow